In inter prediction decoding in H.264, picture data of a current block is decoded by predicting a bi-predictive reference block included in a B slice using, as references, two items of picture data which is data of pictures different from the picture including the current block.
For the H.264 standard, there are motion vector derivation modes available for picture prediction. The modes are referred to as direct modes (see 8.4.1.2.1, 3.45, etc. of NPL 1).
The following two modes of (S) and (T) are available as the direct modes in H.264.
(T): Temporal direct mode (temporal mode). A current block is predicted using a motion vector mvCol of a co-located block (Col_Blk), which is spatially identical to the current block (but temporally different), is scaled by a certain percentage.
(S): Spatial direct mode. A current block is predicted using data on a motion vector (motion data) of a block which is spatially different (but is to be displayed at the same time as a current block).